Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads

[Tigerick]

Wildcat Lake (WCL) Preliminary Specs

Intel Wildcat Lake (WCL) is upcoming mobile SoC replacing ADL-N. WCL consists of 2 tiles: compute tile and PCD tile. It is true single die consists of CPU, GPU and NPU that is fabbed by 18-A process. Last time I checked, PCD tile is fabbed by TSMC N6 process. They are connected through UCIe, not D2D; a first from Intel. Expecting launching in Q2/Computex 2026. In case people don't remember AlderLake-N, I have created a table below to compare the detail specs of ADL-N and WCL. Just for fun, I am throwing LNL and upcoming Mediatek D9500 SoC.

		Intel Alder Lake - N	Intel Wildcat Lake	Intel Lunar Lake		Mediatek D9500
Launch Date		Q1-2023	Q2-2026 ?	Q3-2024		Q3-2025
Model		Intel N300	?	Core Ultra 7 268V		Dimensity 9500 5G
Dies		2	2	2		1
Node		Intel 7 + ?	Intel 18-A + TSMC N6	TSMC N3B + N6		TSMC N3P
CPU		8 E-cores	2 P-core + 4 LP E-cores	4 P-core + 4 LP E-cores		C1 1+3+4
Threads		8	6	8		8
Max Clock		3.8 GHz	?	5 GHz
L3 Cache		6 MB	?	12 MB
TDP		7 W	Fanless ?	17 W		Fanless
Memory		64-bit LPDDR5-4800	64-bit LPDDR5-6800 ?	128-bit LPDDR5X-8533		64-bit LPDDR5X-10667
Size		16 GB	?	32 GB		24 GB ?
Bandwidth			~ 55 GB/s	136 GB/s		85.6 GB/s
GPU		UHD Graphics		Arc 140V		G1 Ultra
EU / Xe		32 EU	2 Xe	8 Xe		12
Max Clock		1.25 GHz		2 GHz
NPU		NA	18 TOPS	48 TOPS		100 TOPS ?

As Hot Chips 34 starting this week, Intel will unveil technical information of upcoming Meteor Lake (MTL) and Arrow Lake (ARL), new generation platform after Raptor Lake. Both MTL and ARL represent new direction which Intel will move to multiple chiplets and combine as one SoC platform.

MTL also represents new compute tile that based on Intel 4 process which is based on EUV lithography, a first from Intel. Intel expects to ship MTL mobile SoC in 2023.

ARL will come after MTL so Intel should be shipping it in 2024, that is what Intel roadmap is telling us. ARL compute tile will be manufactured by Intel 20A process, a first from Intel to use GAA transistors called RibbonFET.

 

[Abwx]

I've actually seen some comparisons between the 7950x, the 13900k, and the M2 Ultra. One of the issues I have with these comparisons is that the M2 is a laptop chip. If the comparisons had, say, a Ryzen 7945HX, I would agree with you, however, comparing Zen 4 or Raptor Lake S to an Apple mobile processor is an Apples to Oranges comparison.

Show me the lower power chips and where they stand.

Here is Ars Technica's chart on low vs. high power chips:

Even that one is suspect because with my own handbrake benchmarks, the 7950X at lower TDPs was MUCH more competitive than in this chart.

Also, the M2 Ultra is uncompetitive in terms of absolute performance. It only wins at perf/watt, so it is clearly a mobile oriented chip.

There s no way that the 7950X efficency would increase that little from 170W to 105W, that s total misleading of the general public for whatever reason.

If we are to look at tests at Computerbase in handbrake perf is 18% lower when TDP is reduced from stock 170W TDP (actually 189W measured in Handbrake) to 88W, that s 1.76x the efficency, at 105W it should be around 1.6x.

Beside as you point it at reduced power level like 105W a 7945HX has even better efficiency at same TDP than a 7950X.

Also the 7950X@88W outperform the 13900K@125W by 8%, so there s no way that the latter could have better perf/watt at 125W than the 7950X@105W, look to me that it s voluntarly biaised numbers, even with a badly set set up it s impossible to yield such grossly fabricated numbers.

Core i9-13900K, i7-13700K & i5-13600K: Gaming-Könige im Test: Benchmarks in Apps

Intel Raptor Lake im Test: Benchmarks in Apps / So wurde getestet / Leistung in Single-Core-Lasten (klassisch)

www.computerbase.de

 

[tamz_msc]

There s no way that the 7950X efficency would increase that little from 170W to 105W, that s total misleading of the general public for whatever reason.

If we are to look at tests at Computerbase in handbrake perf is 18% lower when TDP is reduced from stock 170W TDP (actually 189W measured in Handbrake) to 88W, that s 1.76x the efficency, at 105W it should be around 1.6x.

Beside as you point it at reduced power level like 105W a 7945HX has even better efficiency at same TDP than a 7950X.

Also the 7950X@88W outperform the 13900K@125W by 8%, so there s no way that the latter could have better perf/watt at 125W than the 7950X@105W, look to me that it s voluntarly biaised numbers, even with a badly set set up it s impossible to yield such grossly fabricated numbers.

Core i9-13900K, i7-13700K & i5-13600K: Gaming-Könige im Test: Benchmarks in Apps

Intel Raptor Lake im Test: Benchmarks in Apps / So wurde getestet / Leistung in Single-Core-Lasten (klassisch)

www.computerbase.de

"Grossly fabricated" - only appears so because @eek2121 did not provide the graphs for average power consumption and time taken to complete the encode.

 

[Abwx]

"Grossly fabricated" - only appears so because @eek2121 did not provide the graphs for average power consumption and time taken to complete the encode.

The comparison for the 7950X is at 189W vs 142W, wich explain the only 3% difference in perf , that s what is displayed in this graph as well as in Computerbase test wich include those two power levels, so that s basically an irrelevant graph because theres s not enough power difference to get significant difference in perf/watt.

Beside at stock 170W TDP (actually 189W in Handbrake) the 7950X is 13% faster than the 13900K@253W, and 9% faster than the 13900K@Unlimited power, even at 142W, wich is the 105W (142WPPT) in this graph, it s still faster than the 13900K at 253W and at Unlimited power, wich is not the case in this fabricated graph using faked values.

For accurate comparison one should look at the first serie of Computerbase tests, click on Handbrake and then on +142 Eintrage to display all the tests at different power levels for several CPUs, you ll see the real picture rather than this desperate attempt to make both Intel an Apple look way better than they really are.

Core i9-13900K, i7-13700K & i5-13600K: Gaming-Könige im Test: Benchmarks in Apps

Intel Raptor Lake im Test: Benchmarks in Apps / So wurde getestet / Leistung in Single-Core-Lasten (klassisch)

www.computerbase.de

 

[tamz_msc]

The comparison for the 7950X is at 189W vs 142W, wich explain the only 3% difference in perf , that s what is displayed in this graph as well as in Computerbase test wich include those two power levels, so that s basically an irrelevant graph because theres s not enough power difference to get significant difference in perf/watt.

Beside at stock 170W TDP (actually 189W in Handbrake) the 7950X is 13% faster than the 13900K@253W, and 9% faster than the 13900K@Unlimited power, even at 142W, wich is the 105W (142WPPT) in this graph, it s still faster than the 13900K at 253W and at Unlimited power, wich is not the case in this fabricated graph using faked values.

For accurate comparison one should look at the first serie of Computerbase tests, click on Handbrake and then on +142 Eintrage to display all the tests at different power levels for several CPUs, you ll see the real picture rather than this desperate attempt to make both Intel an Apple look way better than they really are.

Core i9-13900K, i7-13700K & i5-13600K: Gaming-Könige im Test: Benchmarks in Apps

Intel Raptor Lake im Test: Benchmarks in Apps / So wurde getestet / Leistung in Single-Core-Lasten (klassisch)

www.computerbase.de

Oh please, stop your blatant fanboyism. Nearly every test out there shows that the 13900K is faster than the 7950X in its default operation (PL1/PL2 = 253 W) when it comes to short-duration MT benchmarks. I'm not going to threadcrap this thread with past information.

Also encoding speeds heavily depend on the source video. It is entirely possible to use a different source and arrive at different results between benchmarking sites. Your problem is that you believe Computerbase to be the gospel truth, and refuse to acknowledge anything else.

 

[Abwx]

Oh please, stop your blatant fanboyism. Nearly every test out there shows that the 13900K is faster than the 7950X in its default operation (PL1/PL2 = 253 W) when it comes to short-duration MT benchmarks. I'm not going to threadcrap this thread with past information.

Also encoding speeds heavily depend on the source video. It is entirely possible to use a different source and arrive at different results between benchmarking sites. Your problem is that you believe Computerbase to be the gospel truth, and refuse to acknowledge anything else.

Even on the averaged benchmarks at Computerbase the 13900K@253W is just 2% faster than the 7950X@142W, the average is made of 7ZIP, Agitsoft, Blender, CB R15, CB R20, CB R23, Corona 1.3, Digicortex, Handbrake and PovRay.

Notice that they use both CB R20 and R23 wich both favour the 13900K, so you cant say that they rely on AMD friendly benches, not counting that their version of PovRay doesnt use AVX2 for AMD, wich reduce perf of the 7950X by 15%...

Also you should know that the test that please you use the 13900K at 253W and 125W, with such a big 50% lower power it s obvious that there will be a big difference in perf/watt between those two powers, to get a similar comparison with a 7950X they should had set it at 50% lower power as well, that is at 88W PPT like Computerbase did.

So seems that you are good with flawed tests as long as the bias favour your beloved brand, nevermind the scientific accuracy, you can talk of fanboism, you just proved that you re a die hard one.

 

[tamz_msc]

Even on the averaged benchmarks at Computerbase the 13900K@253W is just 2% faster than the 7950X@142W, the average is made of 7ZIP, Agitsoft, Blender, CB R15, CB R20, CB R23, Corona 1.3, Digicortex, Handbrake and PovRay.

Can't you read? I'm talking about short-duration MT benchmarks. By that I mean stuff that usually takes 1-2 minutes to run.

 

[Abwx]

Can't you read? I'm talking about short-duration MT benchmarks. By that I mean stuff that usually takes 1-2 minutes to run.

Since this is about perf/watt it s important to have the TDP under control, i mean if a CPU boost up to 330W for a short duration and that it s enough to perform a bench then the perf/watt should be computed at 330W and not at a base TDP that would be say 250W, that would yield completely fake numbers.

There s also the case where a CPU does half the bench at a boosted power and the other half or so at a base TDP, and that s the same faked numbers if the basis for perf/watt evaluation is the base TDP and not the actual energy over task.

In this respect Computerbase are far better for perf/watt estimations since they state the actual power comsumption and the related perfs, and they do so at a lot of different power points in 10 different benches., FI they tested the 13900K at Unlimited power, 253W, 142W, 125W, 88W, 65W and 45W, and the same for other recent Intel and AMD CPUs.

Last but not least their numbers are way more significant since they use 10 benches, as such any outlier is removed from the equation, there s no way that a CPU would benefit from a bench that is particularly favourable.

 

[tamz_msc]

Since this is about perf/watt it s important to have the TDP under control, i mean if a CPU boost up to 330W for a short duration and that it s enough to perform a bench then the perf/watt should be computed at 330W and not at a base TDP that would be say 250W, that would yield completely fake numbers.

There s also the case where a CPU does half the bench at a boosted power and the other half or so at a base TDP, and that s the same faked numbers if the basis for perf/watt evaluation is the base TDP and not the actual energy over task.

In this respect Computerbase are far better for perf/watt estimations since they state the actual power comsumption and the related perfs, and they do so at a lot of different power points in 10 different benches., FI they tested the 13900K at Unlimited power, 253W, 142W, 125W, 88W, 65W and 45W, and the same for other recent Intel and AMD CPUs.

Last but not least their numbers are way more significant since they use 10 benches, as such any outlier is removed from the equation, there s no way that a CPU wouldl benefit from a bench that is particularly favourable.

The Ars Technica benchmark is about energy efficiency, as in the energy consumed in a standard encoding run.

You're intentionally bulldozing a conversation about energy efficiency, which determines the cost of operation, with talk about perf/watt.

 

[hemedans]

I've actually seen some comparisons between the 7950x, the 13900k, and the M2 Ultra. One of the issues I have with these comparisons is that the M2 is a laptop chip. If the comparisons had, say, a Ryzen 7945HX, I would agree with you, however, comparing Zen 4 or Raptor Lake S to an Apple mobile processor is an Apples to Oranges comparison.

Show me the lower power chips and where they stand.

Here is Ars Technica's chart on low vs. high power chips:

Even that one is suspect because with my own handbrake benchmarks, the 7950X at lower TDPs was MUCH more competitive than in this chart.

Also, the M2 Ultra is uncompetitive in terms of absolute performance. It only wins at perf/watt, so it is clearly a mobile oriented chip.

Atleast in Multithread efficiency gape between M2 max and 7945Hx is just 3%

AMD Ryzen 9 7945HX Analysis - Zen4 Dragon Range is faster and more efficient than Intel Raptor Lake-HX

AMD's new Zen4 CPUs have finally arrived and the new flagship Ryzen 9 7945HX competes with Intel's Raptor Lake HX models. The performance of the new Zen4 cores (Dragon Range) is very good and the Ryzen 9 is also more efficient than Intel's HX CPUs.

www.notebookcheck.net

 

[eek2121]

I don’t mean to interrupt things, but uh, handbrake results will vary significantly based on the source material used and the codec/settings used. This also means power consumption will be wildly different.

This also applies to blender.

 

[Tigerick]

Intel China has released more information about new namings convention as shown here:

• Raptor Lake Refresh - U: Intel Core 3/5/7 Processor 1xxU
• Meteor Lake - U: Intel Core Ultra 5/7 Processor 1xxU with Intel Graphics GT1
• Meteor Lake - H: Intel Core Ultra 5/7/9 Processor 1xxH with Intel Arc Graphics GT2 (dual-channel memory required)

 

[Geddagod]

Intel China has released more information about new namings convention as shown here:

• Raptor Lake Refresh - U: Intel Core 3/5/7 Processor 1xxU
• Meteor Lake - U: Intel Core Ultra 5/7 Processor 1xxU with Intel Graphics GT1
• Meteor Lake - H: Intel Core Ultra 5/7/9 Processor with Intel Arc Graphics GT2 (dual-channel memory required)

Wow this looks horrendous.
Also apparently the "p" series merged into the "h" series

 

[moinmoin]

Intel China has released more information about new namings convention as shown here:

• Raptor Lake Refresh - U: Intel Core 3/5/7 Processor 1xxU
• Meteor Lake - U: Intel Core Ultra 5/7 Processor 1xxU with Intel Graphics GT1

I sure hope there is no overlap in the xx part of 1xxU. Otherwise: yikes!

 

[Tigerick]

Wow this looks horrendous.
Also apparently the "p" series merged into the "h" series

No mention about P series.

The most disappointing is MTL-U comes with GT1 graphics, that is 64EU only: less than Raptor Lake U 96EU. Guess Intel have to use lesser power GPU in order to hit TDP, not to mention the tCPU will remain 2P+8E. So much for ultra naming...

No wonder Intel has to change to monolithic design with LNL-U with 443 configuration: that is 4P+4E with GT3 graphics.

 

[TESKATLIPOKA]

Intel China has released more information about new namings convention as shown here:

• Raptor Lake Refresh - U: Intel Core 3/5/7 Processor 1xxU
• Meteor Lake - U: Intel Core Ultra 5/7 Processor 1xxU with Intel Graphics GT1
• Meteor Lake - H: Intel Core Ultra 5/7/9 Processor 1xxH with Intel Arc Graphics GT2 (dual-channel memory required)

GT2 is 128EU and what about GT1? 64EU?

edit: my question was already answered

edit 2: If It's only 64EU, then I expect much higher IGP clocks compared to Raptor Lake U.

 

[Tigerick]

GT2 is 128EU and what about GT1? 64EU?

edit: my question was already answered

AnandTech Forums: Technology, Hardware, Software, and Deals

Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

www.anandtech.com

Anandtech has full article about it. Go read the article then come back to the slides I shown. The slides are much more informative than Intel US...

 

[igor_kavinski]

So Core 9 reserved for MTL only.

That's funny. Holding back their Refresh parts to prevent them from embarrassing their top MTL SKU

 

[Tigerick]

So Core 9 reserved for MTL only.

That's funny. Holding back their Refresh parts to prevent them from embarrassing their top MTL SKU

There is Core i9 for Raptor Lake Refresh S and HX series but not applicable for new namings convention.

 

[mikk]

No mention about P series.

Looks like the old P is the new H. I guess we will see H parts with 28W as well as 45W.

The most disappointing is MTL-U comes with GT1 graphics, that is 64EU only: less than Raptor Lake U 96EU. Guess Intel have to use lesser power GPU in order to hit TDP, not to mention the tCPU will remain 2P+8E. So much for ultra naming...

Are you saying MTL power consumption goes up? Power consumption is more about clock speed+voltage. If they can go with 96 EUs on Raptor-U they could easily go with 128 EUs on MTL-U. A low clocked 128EU is certainly less power constrained than a high clocked 64EU version. It's more about volume/size/yields/costs....

No wonder Intel has to change to monolithic design with LNL-U with 443 configuration: that is 4P+4E with GT3 graphics.

GT3 for LNL, where did you read this? I would be surprised if it's more than a GT2 128EU. 4+4 is a nice balance though, I prefer it over 2+8.

 

[jpiniero]

AnandTech Forums: Technology, Hardware, Software, and Deals

Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

www.anandtech.com

Anandtech has full article about it. Go read the article then come back to the slides I shown. The slides are much more informative than Intel US...

"Available in your home in 2024, Intel Ultra 64"

 

[aigomorla]

How does china get all the leaks before the US?
And aren't these chips included in the embargo?
Im pretty sure BiliBili is China, and not Taiwan.

 

[TESKATLIPOKA]

Are you saying MTL power consumption goes up? Power consumption is more about clock speed+voltage. If they can go with 96 EUs on Raptor-U they could easily go with 128 EUs on MTL-U. A low clocked 128EU is certainly less power constrained than a high clocked 64EU version. It's more about volume/size/yields/costs....

Raptor Lake U has only 1.3GHz turbo max, so with 50% higher turbo It should perform similarly or better considering It has better architecture.
The question is If It can clock that high with that power budget.
Of course, you are right that on the same process and same architecture, a 1.3GHz 128EU IGP would consume less than a 2.6GHz 64EU IGP.
As you said, 64EU IGP cost less to produce than a 128EU version.

GT3 for LNL, where did you read this? I would be surprised if it's more than a GT2 128EU. 4+4 is a nice balance though, I prefer it over 2+8.

GT3(192EU?) for a CPU config on par with 6 big cores looks kinda overkill, but LNL will be released likely 2-3 years later.
What I find surprising is that there will be such a low configuration. I would say 4+8 would be the bare minimum, but maybe this is a cutdown version.

4+4 would certainly have better performance than 2+8.

Based on this Alder Lake comparison, I made a table to compare 4+4 vs 2+8.
Keep in mind that It shows Alder Lake IPC and It has "unlimited" power budget to keep clocks at the same level.
P-core 5.2GHz: 123 points
P-core 3.9GHz: 100 points
E-core 3.9GHz: 70 points
HT 5.2GHz: 38 points
HT 3.9GHz: 31 points

Threads in use	4P+4E 3.9GHz for both P/E cores	2P+8E 3.9GHz for both P/E cores	4P+4E 5.2GHz P-core 3.9GHz E-core	2P+8E 5.2GHz P-core 3.9GHz E-core
1	100	100	123	123
2	200	200	246	246
3	300 [+11%]	270	369 [+16.8%]	316
4	400 [+17.6%]	340	492 [+27.5%]	386
5	470 [+14.6%]	410	562 [+23.2%]	456
6	540 [+12.5%]	480	632 [+20.2%]	526
7	610 [+11%]	550	702 [+17.8%]	596
8	680 [+9.7%]	620	772 [+15.9%]	666
9	711 [+3%]	690	810 [+10.1%]	736
10	742	760 [+2.4%]	848 [+5.2%]	806
11	773	791 [+2.3%]	886 [+5%]	844
12	804	822 [+2.2%]	924 [+4.8%]	882

Alder Lake 4P+4E is physically larger than 2P+8E, but It will be faster. Considering P-cores are always clocked higher than E-cores, I would say 4P+E4 would end up faster everywhere. Not a bad tradeoff. LNL could look similarly depending on how big of a difference will be between P and E-cores.

 

[Exist50]

Are you saying MTL power consumption goes up? Power consumption is more about clock speed+voltage. If they can go with 96 EUs on Raptor-U they could easily go with 128 EUs on MTL-U. A low clocked 128EU is certainly less power constrained than a high clocked 64EU version. It's more about volume/size/yields/costs....

I agree with your premise that a 128EU part is feasible, and that its absence is more of a product positioning choice, but there're some pretty large differences between gen 12.0 and gen 12.7 uarchs. Might be difficult to get a straight comparison.

GT3(192EU?) for a CPU config on par with 6 big cores looks kinda overkill, but LNL will be released likely 2-3 years later.

Lunar Lake is supposed to be a 2024 product. As in, if they don't hit post-silicon delays, it will launch in 2024, full stop.

Alder Lake 4P+4E is physically larger than 2P+8E, but It will be faster. Considering P-cores are always clocked higher than E-cores, I would say 4P+E4 would end up faster everywhere. Not a bad tradeoff. LNL could look similarly depending on how big of a difference will be between P and E-cores.

The effort is admirable, but there's one core detail that's going to throw off your numbers significantly. Should start seeing hints of that over the next few months.

 

[TESKATLIPOKA]

Lunar Lake is supposed to be a 2024 product. As in, if they don't hit post-silicon delays, it will launch in 2024, full stop.

Lunar lake is a next year product? A year after MTL?
Then next year Arrow Lake is only a desktop product, right?

The effort is admirable, but there's one core detail that's going to throw off your numbers significantly. Should start seeing hints of that over the next few months.

I know what I did is basically Alder or Raptor Lake 4P+4E, didn't have anything better to use, still it was an interesting comparison.
I am very interested in that detail and in what way It trows off my numbers.

 

[Exist50]

Lunar lake is a next year product? A year after MTL?

Yes, more or less.

Then next year Arrow Lake is only a desktop product, right?

I'm not sure the specifics of when each ARL line will launch, but mobile ARL has been widely reported on. ARL and LNL should be contemporaries, for the most part.

I am very interested in that detail and in what way It trows off my numbers.

All in good time